Scene detection methods for MPEG-encoded video signals

This paper evaluates the performance of three scene cut detection algorithms for MPEG-coded video data. A large amount of data taken from actual video sequences is presented to evaluate the scene results. For each method, the percentage of false, as well as failed, scene detection is used as an indication of the performance of the method. The effect of the selected MPEG bit rate and resolution on the performance of the methods are further investigated, to evaluate the methods´ robustness, especially in low bit rate applications. Conclusions on the appropriateness of the methodologies used in the framework of the targeted applications are finally derived.